bangor business school working paper€¦ · bangor business school working paper (cla 24)...
TRANSCRIPT
1
Bangor Business School
Working Paper (Cla 24)
BBSWP/--/---
The relative influence of price and choice factors on retail deposit
quantities
(Bold, Clarendon 12, capitals)
By
John K. Ashton
Bangor Business School
Bangor University, Bangor
Andros Gregoriou
Hull Business School
Hull University, Hull
&
Jerome V. Healy
Hull Business School
Hull University, Hull
October, 2013
Bangor Business School
Bangor University
Hen Goleg
College Road
Bangor
Gwynedd LL57 2DG
United Kingdom
Tel: +44 (0) 1248 382277
E-mail: [email protected]
2
The relative influence of price and choice factors on retail deposit quantities
Abstract
This study explores how price and non-price factors influence the quantity of retail deposits
held by depository institutions. Price factors examined include the policy or base rate and
retail deposit interest rates set by individual banks, and non-price factors include the branch
network and the number of deposit accounts offered by individual banks. The analysis is
undertaken for a sample of UK building societies over twenty three years using a
disaggregated data set of instant access retail deposit accounts. The methodology uses a two
stage econometric procedure involving system estimators in a panel framework using SUR
and GMM to control for contemporaneous correlation and endogeneity concerns. Both price
and non-price factors are significant influences of retail deposit quantities. We conclude risk
assessment of retail deposit quantity and monetary policy transmission would benefit from
considering price and non-price factors, rather than only price factors.
3
The relative influence of price and choice factors on retail deposit quantities?
1. Introduction
This study examines the effect of price and non-price factors on the quantity of retail deposits held by
depository institutions. Specifically we test the relative influence of price factors, such as depository institution
specific deposit interest rates, relative to non-price factors including the scale of the branch network and the
number of retail deposit accounts offered to depositors’, over changes to retail deposit quantity. The study
employs a disaggregated dataset of deposit quantities, instant access deposit accounts and interest rates offered
by UK building societies between 1989 and 2011 to test for the relative influence of these factors. To overcome
contemporaneous correlated and jointly determined factors and endogenity concerns both Seemingly Unrelated
Regression (SUR) and a Generalised Method of Moments (GMM) panel models are employed. It is reported
depository institution specific interest rates and the number of deposits accounts marketed are both significant
influences on deposit quantities. It is concluded methods used to estimate changes in deposit quantity or assess
the impact of policy or base rate changes would benefit from considering both price and non-price factors.
Determining what factors influence the quantity of retail deposits held by depository institutions has a
considerable economic importance. ‘Core’ deposits are a stable form of financing depository institutions (e.g.
Cornett, McNutt, Strahan and Tehranian, 2011; Huang and Ratnovski, 2009; Shin 2009; Sheehan, 2013) and
higher deposit levels as a proportion of all funding are significantly associated with increased lending (Cornett
et al. 2011). In this role deposits confer relative benefits over wholesale finance, as depositors have fewer
incentives to monitor banks and withdraw funds based on negative public signals. Retail deposits are therefore a
relatively robust and less febrile source of depository institution financing relative to wholesale funding1 and
since the 2007 financial crisis an increasingly cost effective form of funding. These benefits have received
institutional recognition within prudential regulation through the definition of stable and less stable deposits by
the Basel III agreement (Basel Committee on Banking Supervision 2010 [hereafter Basel]).
Despite the importance of retail deposits it is not always clear what depository institution actions
influence their quantity. When addressing this question, conventional wisdom suggests price factors such as
wholesale costs, base or policy rates, and bank specific interest rates are highly significant influences. It is
widely assumed (see Hutchingson and Pennacchi, 1996) the wholesale or market rate will negatively influence
1 Low levels of deposit finance has been directly associated with actual banking failures, such as the Northern Rock in the UK in 2007,
where the bank held only 23% of its liabilities in the form of deposits (Shin 2009) funding most of its growth through interbank lending and
securitisations until mid-September 2007 (National Audit Office 2009).
4
deposit quantities. If the risk free rate on wholesale investments increases, deposits appear less attractive and
depositors shift funds from retail deposits to other investments. Alternatively, subject to the accessibility and
viability of banks, depositors observing a higher rate of retail deposit interest, switch funds from other
investments into retail deposits. The depository institution will also adjust retail deposit interest rates to
accommodate changes in the wholesale cost of funds and also to attract deposit funding when required for bank
specific reasons. Emphasising price as the key influence of deposit quantity is therefore widely adopted in
banking regulation such as within the Basle regulations on bank interest and liquidity rate risk (see Entrop,
Wilkens and Zeisler 2009; Basel, 2010; Sheehan, 2013).
The assumption that price is the key determinant of deposit quantity is also common in an allied and
associated literature examining the transmission of monetary policy decisions. This research has extensively
examined how base or policy rate changes are transmitted to retail interest rates, where such interest rates are the
key influence ultimately affecting deposit quantities. This literature reports retail interest rate setting is
‘sluggish’, ‘sticky’, or lagged (e.g. Hannan and Berger 1991; Rosen 2002; De Graeve, De Jonge and Vander
Vennet 2007; Fuertes and Heffernan 2009) in many financial services and particularly deposit markets.
Therefore factors other than price may be influencing this process. This is a plausible perspective in light of the
rising deposit funding requirements of many banks and historically low base or policy rates seen internationally
since the 2007-2009 financial crisis. Indeed within such a low interest rate period, the ability for banks to
compete for deposits in terms of interest rates alone is constrained and it is not implausible to suggest other non-
price factors are influential in these markets.
In light of the foregoing this study provides an academic contribution through assessing liquidity and risk
management and monetary policy transmission. This is undertaken by assessing whether non-price as well as
price factors are influential determinants of deposit change. To achieve these aims the study is divided into five
sections. After this introduction a review of how depository institutions can influence the quantity of retail
deposits is provided. In section three the empirical procedures adopted in the paper are outlined, including the
hypotheses to be tested, the data sets employed and the testing procedures. In section four, the empirical results
are discussed and in section five, conclusions are drawn and suggestions for future work indicated.
5
2. Literature review
Literatures examining the link between depository institutions’ actions and retail deposit quantities have
developed around a number of themes. Within this review two such areas are explored. First, the implications of
different levels of retail deposits for the viability of depository institutions and the assessment of their liability
and liquidity risks. This literature suggests that factors other than wholesale cost of funds and the prevailing
interest rate paid on individual deposit accounts are important. Second, this work has implications for the
transmission of changes in policy interest rates from central banks to depository institutions’ balance sheets.
Assessing the speed of this process is an essential step in quantifying the effectiveness of monetary policy
actions in influencing the real economy. Academic attention in this literature has focused on the relationship
between official rates of interest or the wholesale cost of funds, and the interest rates set by individual
depository institutions for banking services. These two literatures are examined in turn2.
Assessing the determinants of deposit quantities and how deposit markets operate competitively has long
been associated with the modelling of liability risks posed by demand deposits and the appropriate methods for
hedging such positions (see Cornett et al. 2011; Huang and Ratnovski 2009). These academic literatures have
considered the market cost of funds and further factors influencing retail deposit quantities. For example
Penacchi and Hutchinson (1998) considered the importance of market power in influencing interest rates and
interest rate risk. Jarrow and van Deventer (1998) developed the plausibility of how demand deposits are
employed by depository institutions and depositors, through using an arbitrage free procedure acknowledging
depository institutions can’t buy deposits, and depositors contribute to yet do not issue deposits. Kalkbrener and
Willing (2004) considered the term structure of liquidity to enable the forecasting of future deposit quantities.
Further Nyström (2008) acknowledges deposit account volumes are closely related to other financial services. It
is assumed deposits are often drawn from transactional or current accounts and deposit accounts are then
subsequently used to amass money before further investments or spending is undertaken.
2 This examination of the determinants of deposit supply from depository institution or market actions on the cost of funds in how deposit
services are priced or are presented to customers is distinct from literatures examining influences on the deposit supply more widely. Literatures examining determinants of savings rates internationally is extensive and has proposed a variety of factors associated with
increased savings rates linked with customer actions and behaviours and wider social and economic institutions. Key developments in this
literature have included, normative theories of what savings should be undertaken (Hall 1978), the composition of individual portfolios (e.g. Huberman and Jiang 2006), survey work as to which types of consumers actually make poor financial decisions and quantifying the
magnitude of current low levels of saving (Banks and Tanner 1999). Indeed a wide range of factors are associated with different level of
saving including for example the availability of credit, retirement age and possible intergenerational transfers (Kirsanova and Sefton 2007), different attitudinal factors (e.g. Canova, Rattazzi, and Webley 2005), the level of economic development (Deaton and Paxson 2000) and the
consumption life cycle (Browning and Crossley 2001). While this research has done much to extend our current comprehension of why
individuals and families choose to save more or less, the literatures examining what strategies banks may undertake to influence the supply of deposits is far less extensive, at least in quantity and for the purposes of this discussion is a distinct area of investigation.
6
To summarise, two main findings emerge from this literature. First, it is generally acknowledged that
comparative assessment of distinct influences on deposit quantities is absent. Second, it is considered that
regulatory authorities such as the Basel Committee on Banking Supervision have not acknowledged the
influence of factors other than the wholesale cost of funds in influencing the supply of retail deposits.
An extensive literature has also examined how market or official rates of interest are transmitted to the
interest rates of retail deposits and other financial services. This literature assesses how bank interest rate setting
is influenced by changes in the policy or base rate or the wholesale cost of funds, indicating how such monetary
policy actions can eventually influence the supply of deposits. This approach assumes, not implausibly, that
depository institutions specific interest rates are a key determinant of deposit quantities. Evidence of this and
other factors is presented in many studies which have identified the competitiveness of retail financial services
markets (Calem and Mester 1995; Heffernan 1997); interest rate asymmetry (De Haan and Sterken 2004), the
structure of the banking industry, both in the US and Europe (Corvoisier and Groop 2001; De Graeve et al.
2004; Hannan and Berger 1991), lending channel effects (De Graeve et al. 2004), bank efficiency (Fuertes and
Heffernan 2009), macroeconomic changes (Gambacorta 2008), regulation (Chong 2010) and the scale of official
or base rate changes (Fuertes, Heffernan and Kalotychou 2010) as influential factors influencing the relationship
between official or market rates and retail interest rates.
In summary, it is clear that the link between policy rates or the wholesale cost of funds and interest rate
setting by individual depository institutions is lagged, at least in the short-run. Despite the many insights offered
by this literature the reasons for these rigidities are still not fully explained. These rigidities are at least
implicitly acknowledged in the Basel III agreement where retail deposits are subdivided into ‘stable’ deposits,
where depositors have long standing relationships with depository institutions, and ‘less stable’ retail deposits,
more sensitive to deposit interest rates, offered over alternative, non-branch distribution channels and used by
more ‘sophisticated’ or high net worth individuals (Basel, 2010). While this literature has mainly focused on the
transmission process from base rate changes to retail interest rates, the link between both policy rate changes,
retail interest rates, and changes in retail deposit quantities remains relatively under researched.
3. Hypothesises, Data and Analytical Framework
This section summarises the empirical assessment. Initially, the hypotheses drawn from the preceding literature
review are outlined. Next, the data is described, and finally the empirical design is elaborated.
7
3.1 Hypothesises
Within the preceding review it was articulated that while many depository institution actions could influence
deposit quantities, price or interest rates are widely considered to be a key determinant. Therefore depository
institutions’ specific deposit account interest rates observed in each year (t) for each depository institution (i) are
considered in combination with base rates. It is expected that depository institution specific interest rates will
have a positive relationship with deposit quantity change, subject to the influence of base rates i.e. higher
deposit returns will attract higher deposit quantities. Models are estimated using either, the average or maximum
interest rates offered in any year by each depository institution, reflecting different ways in which competition
could operate in deposit markets. If the interest rate on all deposit accounts offered by a particular depository
institution informs depositor decisions then an average level of interest is appropriate to consider. Alternatively,
a depository institution could compete by offering accounts with higher interest rates to attract new depositors,
where new depositors choose the deposit account with the highest available interest rate. This competitive
process (see Klemperer 1995 and Carbo-Valverde, Rodriguez-Fernandez and Hannan 2011) suggests the
maximum interest rate is a more appropriate measure to determine the influence of interest rates on deposit
quantities. We therefore examine:
Hypothesis 1. Deposit quantity change is positively associated with depository institutions’
specific, deposit account interest rates, be these average or maximum interest rates.
The second hypothesis is that non-price factors also influence deposit quantities. Both the scale of the
branch network and the level of deposit account choice are employed as non-price factors. The scale of the
branch network has been previously observed to be a significant attraction to depositors (see Kiser 2002) and
through a wider branch network a depository institution can become more convenient for a depositor to access.
The number of deposit accounts offered by a depository institution has also been viewed to be a significant
influence on purchase decisions, although the expected relationship between the degree of choice and deposit
quantities is not clear cut, as Kamenica (2008) reports. While customers’ benefiting from more rather than less
choice is one of the simplest choice theoretic principles, there have been many reported cases where this
principle is violated in financial services and other markets (see Iyengar and Lepper 2000; Boatwright and
Nunes 2001; Huberman and Jiang 2006; Carlin and Manso 2010). Thus, while depository institutions are
expected to offer more deposit accounts in the hope of attracting more depositors’ funds, it is acknowledged this
may not always be the outcome. Specifically, since financial services markets are characterised by poor
8
customer comprehension (Bucks and Pence 2008, FSA 2006), and individuals face difficulties when
comprehending monetary values and yields accurately (e.g. Butler and Loomes 1988; Shafir, Diamond, Tversky
1997), too many choices may create depositor confusion and actually reduce deposit quantities. We therefore
examine:
Hypothesis 2. Deposit quantity change is associated with the scale of the branch network and/or
the number of deposit accounts offered by an individual bank.
Lastly, it is also acknowledged that there are further bank specific factors which may also influence
deposit quantity change. We therefore also consider whether a depository institution has merged with another
depository institution during each year, the percentage of new deposits obtained which are retail instant access
deposits relative to all of the depository institution’s deposits. This last factor is important if the depository
institution has become reliant on funding from sources other than instant retail deposits, such as notice deposits.
3.2 Data
The choice of data employed is complicated due to depository institutions organisational structures. Many
depository institutions operate through holding company structures and use a variety of subsidiaries through
which deposited funds are often reallocated and transferred (see Cremers, Huang and Sautner 2010). This makes
it challenging to determine where retail deposits are derived from within a depository institution and link the
source of deposits to a particular deposit account and interest rate. Indeed the use of such organisational
structures has potentially constrained research examining the relationship between deposit quantities and bank
specific interest rates. To accommodate these concerns a relatively small and homogeneous sample of
depository institutions are considered: UK building societies. These mutually owned institutions are often small,
have a relatively simple organisational structure and do not operate multiple subsidiaries. Further due to both
their relatively limited scale and legal restrictions most of these banks have been unable to receive external
financing during the sample period. These features therefore make this form of depository institution
particularly suited for this form of assessment due to their limited organisational complexity. This said we
acknowledge that inference needs to reflect that this sample of depository institution has distinct features to
other often larger commercial retail banks providing deposit services.
The geographic market considered is the entire United Kingdom. While there is evidence that the UK
displays limited features of separate geographical banking markets (Ashton, 2001), this approach is consistent
9
with both past academic and regulatory assessments of this market (e.g. Competition Commission 2002; Fuertes
and Heffernan 2009; Heffernan 2002). The product market considered is the instant access retail deposit account
market. These accounts are distributed through both branch networks and also over the internet, by post, and
using telephone networks. While it is acknowledged that branch and remotely distributed accounts could be
separate product markets, past research has reported retail deposit interest rate setting does not significantly vary
between these forms of distribution (Ashton 2002). Deposit accounts which require notice before withdrawal,
impose penalties for immediate withdrawal or receive tax exemptions are excluded due to their differing product
terms and conditions.
The deposit account interest rate data and the number of deposit accounts offered by each bank in each
year was constructed using data issued by the Moneyfacts magazine; this data has also been employed within
regulatory assessments of the banking industry (e.g. Competition Commission 2002) and past academic work
(e.g. Ashton and Hudson 2008; Heffernan 2002; Fuertes and Heffernan 2009). This data includes all instant
access retail deposit accounts recorded for the sample banks from January 1989 to December 2011. Deposit
account features such as access to a cash card are not considered in the assessment. While these features have a
real interest rate value (Ashton and Letza 2003) their measurement is problematic (Heffernan 2002).
Throughout, the building society specific interest rates considered are for £5,000 to £9,999 deposited. This
choice of interest rate tier or interval is made to reflect average customer deposit sizes. The Family Resource
Survey, which provides estimates of UK individuals financing decisions, confirms that for the period 1995 to
2011 (for which data could be obtained), the average deposit, for persons with savings, is between £3,000 and
£8,000. This range is consistent with the size of deposit accounts considered in this study (an average instant
access deposit of £4,684)3. This data is used to provide both average and maximum interest rates for instant
access deposit accounts for each building society annually. The UK policy or base rate is used to quantify the
cost of funds. While the London Inter-Bank Offer Rate (LIBOR) has been employed in past studies, concerns
exist as to its determination (Abrantes-Metz, Kraten, Metz and Seow 2012) and how well this interest rate
actually reflects official monetary policy changes (Fuertes et al. 2010).
Data on deposit quantities and branch numbers is taken from annual regulatory accounts obtained from
the Building Society Association. These regulatory accounts offer relatively more detail on the source and
classification of deposits compared to annual reports and accounts provided by proprietary banks. Using these
3 Data on depositors and the average deposits held by depositors using each building society was only obtained for some and not all years.
Data for this estimate includes all sample building societies between 1989-99, 2005-06 and 2008-11.
10
regulatory accounts, instant access deposit quantities can be directly linked to the average level of building
society specific interest rates offered for instant deposit accounts. The frequency of the data is assumed to be
annual, reflecting the lowest frequency of data in this assessment. Whether a building society has merged in a
particular year was also recorded as a control on deposit change; this variable was defined using the information
drawn from the BSA (Building Society Association) year book (2013), within which all merger events over the
sample period are recorded. In all cases the mergers undertaken have been with other building societies; when a
merger has been undertaken with a commercial bank, the building society loses its mutual status (and its’
relative organisation simplicity) and is excluded from the analysis. A summary of the variables used in the
model is provided in Table 1.
Table 1. Variable definition
Variable name Variable description Data source
Δ deposits*i, t+1 The change in the quantity of instant access
retail deposits received over a year.
Annual regulatory accounts from
the Building Societies Association
and Annual reports and accounts
of individual societies.
Average Bank Interest
Rate
The average instant access interest rate on all
such deposit accounts offered by the bank in the
particular year. It is assumed £5,000 deposited
is an average deposit.
MoneyFacts PLC
Maximum Bank
Interest Rate
The maximum instant access interest rate on all
such deposit accounts offered by the bank in the
particular year. It is assumed £5,000 deposited
is an average deposit.
MoneyFacts PLC
Base Rate The policy, base or official interest rate. Bank of England
Number of deposit
accounts
The number of deposit accounts records the
different instant access deposit account products
offered by an individual bank in any year.
MoneyFacts PLC
Number of Branches The number of branches in a building societies
branch network. This includes only full
branches and excludes for example associated
estate agencies etc.
Annual regulatory accounts and
various Building Society
Association BSA year books.
Bank Merger Whether a bank has merged in a particular year
or not with another building society.
Information on building society
mergers from the BSA (Building
Society Association) year book
(2013)
Proportion of funding
from instant access
deposits
The quantity of deposits which are instant
access divided by total building society
deposits.
Annual regulatory accounts and
accounts of individual societies.
While this data set brings considerable advantages in terms of the simplicity of the organisations, markets
and data assessed, there are also draw backs. The sample building societies are relatively small, are mutually
owned, and potentially employ non-price forms of competition to a lesser extent than larger banks with more
developed marketing functions. Also the overall dataset is small reflecting the challenges of linking deposit
11
quantity data accurately with retail interest rates in a meaningful manner. Moreover, within the sample period
many building societies have operated in the UK, of which only a proportion are considered. This occurs as
many of these building societies have merged and/or demutualised during the sample period. Also some
building societies have incomplete interest rate or accounting data. For example two building societies altered
how they reported deposits over the sample period and were subsequently dropped from the analysis. To avoid
concerns with survivorship bias a sample of 26 building societies are considered for which data is available for
at least 19 years. The average duration of a building society in the dataset is 21.8 years. The 26 building
societies examined are listed in Appendix 1.
3.3 The Analytical Framework
To test the hypotheses a two part testing methodology is employed. This empirical design is influenced by
challenges in quantifying the research questions and the data format. First, building societies and deposit
quantities are exposed to similar systematic shocks over the sample period. This could lead to cross-building
society residuals being contemporaneously correlated. Secondly, the variables considered are susceptible to
endogeneity. For example setting retail interest rates or the number of branches could both influence the
quantity of deposits and be actions triggered by a shortage or over-abundance of deposit funds. Such
endogeneity can bias econometric results and needs to be addressed. Third, the influence of different price and
non-price bank actions on deposit quantities may be interdependent and jointly determined. For example a
building society requiring higher deposit funding may decide to offer more deposit accounts and provide higher
interest rates to encourage depositors to use the building society, either distinctly or in combination. Lastly, the
data considered could be influenced by further time series and cross-sectional dependencies not identified or
included in the model. Accordingly concerns with contemporaneously correlated variables, endogeneity
concerns, joint determination, possible unidentified data dependencies and a desire to test between explanatory
variables leads to the following econometric procedures being adopted.
The models to be estimated are written as:
titititi
titititititideposits
,,,,,2
,,,,,1,,
deposits short term from funding of ProportionMergers
RatePolicy rateinterest society building Average
(1)
titititi
titititi
Values FittedBranches of Number
Products of Numberdeposits
,,,,,2
,,,1,,
][
(2)
12
where i represents the building societies and t denotes the annual time period; i captures the time-invariant
unobserved building society-specific fixed effects, and the t captures the unobservable individual-invariant
time effects. We also re-estimate (1) with average building society interest rate replaced by maximum interest
rates. The fitted values of deposit change from model (1) are included in model (2), and then and vice versa.
Thus the two equations (3) and (4) can be estimated as panel data models4. As before, these latter two models
are estimated using maximum interest rates as well as average bank interest rates.
tititi
tititi
titititititi
deposits
deposits
,
*
,,,3
,,,,2
,,,,,1,,
deposits short term from funding of ProportionMergers
RatePolicy rateinterest society building Average
(3)
tititi
titititititi
deposits
Branches of NumberProducts of Numberdeposits
,
*
,,,3
,,,2,,,1,,
(4)
Through capturing the effect of time varying fixed effects, further influences on the data can be
captured in the estimation. The fixed and time effects are significant in all cases suggesting that the building
society and time-specific shocks differ significantly across the building societies in our sample. A test for first
order serial correlation is also undertaken and is insignificant suggesting that the panels do not suffer from serial
correlation. The Jarque-Bera normality test is also undertaken to determine if the residuals of the models are
normally distributed. These tests imply that the empirical estimates obtained are not due to any outliers in the
data.
To accommodate joint determination of the explanatory factors, the model is estimated using a GMM
system of equations in first differences and levels. This system estimator combines the standard set of
transformed equations in first differences with an additional set of equations in levels. The first set of
transformed equations uses the lag levels as instruments and the level equation uses the lagged first differences
as instruments. Their validity is based on the following two moment conditions:5
,
,
( )0 z 1,
( )
it it i t z
it it i t z
a e DepositsE for
a e W
(5)
4 We conduct a Hausman test to determine if the explanatory variables are endogenously determined. This will also determine the choice of
panel estimator for our econometric analysis. 5 The time-varying matrix of instruments for the first difference GMM estimator can be observed in Blundell and Bond (1998).
13
where W denotes the explanatory variables in (1) and z represents the lag structure of the GMM estimator. The
composite error term displayed in equation 5 is a result of the system error of the estimation of forward looking
expectations obtained from the GMM System estimator model in a panel framework (see Ghosh and Gregoriou
2008). In addition to dealing with explanatory variables being jointly determined, the GMM system has the
added advantage of reducing the poor precision of the GMM single equation estimators.
The GMM single equation estimator is employed instead of a Two Stage Least Squares method
because the GMM is asymptotically efficient under non-restrictive assumptions about error autocorrelation and
heteroscedasticity (Biorn and Klette 1999). The validity of the instruments is examined using a Sargan test
under the null hypothesis that the instruments used are valid. This test reports a p-value of zero confirming that
the instruments used are not valid; this indicates subsequent empirical findings based on this estimator would be
weakened and the estimated coefficients suffer from poor precision (see Staiger and Stock 1997). A possible
reason for these weak instruments is that the time dimensions of the panels are relatively small (23 annual
observations) and implies there is a weak correlation between the regressors and the instruments. This problem
is addressed by using the panel GMM system estimator proposed by Blundell and Bond (1998) which radically
reduces the imprecision associated with the single equation estimator. The Sargan test confirms the validity of
the instruments in both GMM system models with this approach. 6
After equations 1) and 2) are estimated separately for each individual building society, we examine if
the cross-building society residuals are contemporaneously correlated with the use of LM tests. To examine
whether cross- building society residuals are contemporaneously correlated the Breusch and Pagan (1980)
Lagrange Multiplier (LM), statistic is undertaken:
, (6)
where is the squared ijth
correlation coefficient of cross- building society residuals. Under the null of no
contemporaneous error correlations across the building society, the test statistic is asymptotically distributed
with N(N-1)/2 degrees of freedom, where N denotes the number of building society in the panel. The p-value of
6 The Three Stage Least Squares (3SLS) panel estimator also estimates a system of equations simultaneously and is regarded as an
alternative to the GMM system estimator. However, we implement the GMM system estimator, given that it accommodates for the
possibility of joint determination of an equation system with different instruments for different equations (Schmidt 1990).
LM
12
2 1
n i
LM ij
i j
T r
2
ijr
2
14
the LM test statistic is zero, which rejects the null hypothesis, suggesting that error series are
contemporaneously correlated across all the building society. To accommodate this concern a SUR econometric
methodology is also employed for the empirical analysis.
Even though the SUR estimator takes into account contemporaneous correlation across building
society, it fails to capture the second concern, that of endogeneity in the explanatory variables of the panel. In
order to formally test the explanatory variables for endogeneity, a Hausman test for the hypothesis that the
explanatory variables are strictly exogenous is performed. If the null hypothesis is rejected, it leads to the
conclusion that the explanatory variables in equation (1) are endogenously determined. The Hausman test rejects
the null hypothesis at all conventional significance levels. Endogeneity for the explanatory variables7 is
therefore addressed using the single equation GMM panel estimator developed by Arellano and Bond (1991).
4. Empirical Results
In this section descriptive statistics of the building society interest rates, deposit quantities, control variables and
numbers of deposit accounts and branches are presented in Tables 2 and 3. The results from the two equations 3)
and 4) are presented in Tables 4 and 5 for averaged and maximum deposit interest rates respectively.
The descriptive statistics are considered overall and also for four different 5 or 6 year periods to
illustrate how variables have altered over time. It is reported that while the number of building society
considered over time remains constant, the scale of the other variables rises significantly. The quantity of instant
access retail deposits increases by over 73% over the sample period. The percentage change in instant access
deposits for individual building society is also highly variable with extremes of change ranging from an 81%
increase and a 35% decrease. Such large changes are generally associated with merger events, troubled building
society and changes in the level of instant access deposits collected as a proportion of all deposits. The number
of instant access deposit accounts offered by each building society increases from an average of 1 in 1989 to 6.7
in 2011. The number of deposit account products also varies between building society, increasingly greatly as
the base rate has fallen. The returns on deposit accounts have also fallen reflecting the reduction in the base rate
over time. There is also considerable difference between the average and maximum interest rate with an average
difference of 1.1% interest. The size of branch networks gradually displays variation throughout the sample
7 The results of the Hausman test are not reported by the authors, but are available upon request.
15
period, falling between 1989 and 2006, before a small increase in the last time period (2007-2011). The
percentage of deposits which are instant access also varies over time displaying a gradually decline.
Table 2: Descriptive Statistics of Deposit Supply, Deposit Interest rates and Deposit Account Choice.
Overall 1989-1994 1995-2000 2001-2006 2007-2011
Number of observations 584 154 156 156 118
Number of building societies 26 26 26 26 26
Annual average base rate 6.01 10.03 6.26 4.42 2.55
Average retail Interest rate 3.87 7.18 3.64 2.58 2.09
Maximum retail interest rate 5.01 8.22 4.66 3.75 3.49
Average % of merger observations 2.23 1.95 1.92 1.28 4.24
Quantity of instant
retail deposits £m
Average 3063.91 1543.93 1543.93 39862.30 5149.05
Minimum 6.66 7.95 6.66 11.95 85.38
Maximum 93511.90 24950.70 39862.30 82092.60 93511.90
% of deposits
which are Instant
access
Average 88.28 93.09 87.96 90.58 80.47
Minimum 32.71 35.63 41.95 60.56 32.71
Maximum 100 100 100 100 100
% change in
instant deposit
quantities
Average 7.23 8.26 8.42 8.70 2.34
Minimum -35.88 -35.88 -12.47 -29.40 -28.41
Maximum 81.20 56.75 61.71 61.46 81.20
Branch network
size
Average 60.18 65.12 57.77 57.44 60.65
Minimum 1 1 1 1 1
Maximum 900 869 693 682 900
Number of instant
account accounts
Average 3.00 1.24 2.06 3.61 5.47
Minimum 1 1 0 0 1
Maximum 28 3 12 21 28
Correlations between model variables are reported in Table 3. The strongest relationships are observed
between the retail interest rates and the base rate. Significant positive correlations are also observed between the
level of retail interest and the number of branches, and between the level of all interest rates and change in the
quantity of instant access deposits. The incidence of mergers and branch numbers is also significantly positively
correlated.
16
Table 3: Correlations between variables
Correlations between variables
Policy
Rate
Average
Average
Bank
Rate
Maximum
Retail
Interest
Rate
Change
in
Deposit
Supply
Branch
Network
Number
of
Deposit
Accounts
% of Merger
Observations
% of
Funding
from
Instant
Access
Deposit
Accounts
Policy Rate
Average 1 0.88* 0.79* 0.29* 0.02 -0.38* 0.02 0.28*
Average
Retail Interest
Rate
1 0.95* 0.21* 0.10* -0.24* 0.02 0.17*
Maximum
Retail Interest
rate
1 0.23* 0.16* -0.08 0.06 0.11*
Change in
Deposit
Supply
1 0.00 -0.03 0.09 0.13
Branch
Network 1 0.06 0.19* -0.01
Number of
Deposit
Accounts
1 0.05 -0.28*
% of Merger
Observations 1 -0.15*
% of Funding
from Instant
Access
Deposit
Accounts
1
* denotes statistically significant at 10% significance.
The estimates of the equations 3) and 4) are provided in Table 4 for average retail instant access
interest rates and in Table 5 for the maximum retail instant access interest rates. Within Table 4, the average
building society interest rate and the number of deposit accounts are both statistically significant and the scale of
branch network is statistically insignificant. The scale of the coefficient values in the two equations are similar
for both the SUR model and the GMM System8. This indicates that both building society specific interest rates
and the number of deposit products issued by a bank influence the quantity of deposits. This provides evidence
supporting Hypothesis 1 – that deposit quantity change is positively associated with average, building society
specific deposit account interest rates. For Hypothesis 2 - deposit quantity change is associated with the number
of deposit accounts offered by an individual bank as significant positive values are recorded, yet not with choice
8 We use a GMM system estimator with one lag structure as reported in equation 5. In empirical estimates of the GMM system estimator it is
not common to report all lags of explanatory variables in both levels and first differences. This is because they are simply used for
estimation purposes and add little value to the equation estimates. This is the case in previous empirical studies that use the GMM system estimator to derive their results. For example see among others Ghosh and Gregoriou (2008).
17
by more or fewer building society branches. Therefore while our prior expectation was that greater choice in
deposit accounts could result in either an increase or a decrease in deposit quantity, increasing the level of
deposit account choice has a positive impact in this market. This indicates concerns with customer confusion do
not dominate and more choice has resulted in greater sums deposited. The relative scale of these influences
indicates that while both deposit interest rates and the number of deposit accounts are significant influences, the
former are marginally more influential on deposit quantities.
The model estimates also provides further information. Two control variables have been included to
accommodate the influence of both mergers and also the proportion of deposits a building society has raised
which are instant access. The merger variable is seen to be statistically significant while the proportion of instant
access deposits is insignificant in both cases indicating notice deposits maybe a distinct product market. In all
model estimates the influence of the base or policy rate is seen to be statistically significant. This finding is
unsurprising in light of the emphasis placed on this variable within the transmission of monetary policy and the
risk assessment of deposit quantities literatures. This said these results indicate that using only the base rate
when predicting deposit quantities will only partially determine deposit quantities, and while this variable
maybe a fair predictor of retail deposit quantities it is clearly not the only meaningful predictor of deposit
change.
In Table 5 the influence of the maximum interest rate offered by a bank in a particular year is
considered. It is observed that for both SUR and GMM models this variable has an insignificant influence on
deposit quantities. This indicates the form of competition operating in the deposit market has a bearing on the
transmission of interest rate change to deposit quantities. For example the low level of switching observed in
deposit markets (see Independent Commission on Banking 2011) may have reduced the possible influence of
maximum interest rates. In common with results from Table 4, the number of deposit accounts has a significant
influence and the number of branches has an insignificant influence on deposit quantity change. Mergers are
also a significant control variable and the percentage of instant access deposits from all deposits is an
insignificant influence. Again the base rate has a large and statistically significant influence on change in the
quantity of instant access deposits a bank receives or loses in a particular year.
18
Table 4. Model Estimates: Average Interest rate
Estimates from Equation 3
Variable GMM System Estimates SUR Estimates
Constant -3.69 (-1.16) -3.02 (-0.93)
Average Bank Interest Rate 0.62 (1.67)* 0.56 (1.35)
Base Rate 1.41 (4.54)** 1.40 (4.48)**
Mergers 7.09 (2.07)** 7.03 (2.06)**
% Instant Access Deposits 0.055 (1.51) 0.052 (1.45)
Δ deposits*i, t+1 -0.73 (-0.37) -0.90 (-0.46)
ai (0.00) (0.00)
bt (0.00) (0.00)
SE 0.55 0.62
AR(1) (0.40) (0.46)
NORM(2) (0.56) (0.62)
Diff Sargan NA (0.67)
Hausman test NA 97.66 (0.00)
R2
0.39 0.30
Observations 565 565
Estimates from Equation 4
Variable GMM System Estimates SUR Estimates
Constant 7.28 (10.47)** 7.13 (10.88)**
Number of Deposit Accounts 0.21 (2.88)** 0.16 (2.66)**
Number of Branches 0.34 (0.98) 0.29 (0.84)
Δ deposits*i, t+1 0.86 (3.22)** 0.72 (2.55)**
ai (0.00) (0.00)
bt (0.00) (0.00)
SE 0.22 0.16
AR(1) (0.22) (0.18)
NORM(2) (0.24) (0.20)
Diff Sargan NA (0.44)
Hausman test NA 98.55 (0.00)
R2
0.22 0.16
Observations 565 565
Notes: AR(1) is the first order Lagrange Multiplier test for residual serial correlation, undertaken on the
residuals for the SUR estimates and on the first difference of the residuals for the GMM system because of the
first difference transformations involved. SE represents the standard error of the panel estimator. ai and bt are
the fixed and time effects. Sargan tests follow a distribution with r degrees of freedom under the null
hypothesis of valid instruments, for the four instruments in our empirical model. Note: the Difference-Sargan
test is applicable to the GMM system estimator due to the first difference transformations involved. To
establish the validity of the instrument set. NORM(2) is the Jarque-Bera normality test. The endogenous
explanatory variables in the panel are GMM instrumented setting, (.) are p values, (.) are t statistics, *
indicate significant at the 10% level and ** indicates significant at the 5% level
While all estimated models have acceptable diagnostic test results, the degree of model fit reported in
the models indicates there is still much variation in the data which is not fully explained by these models.
Therefore both the transmission of policy or base rate changes to deposit quantities and the risk assessment of
deposit quantities could prudently employ a wider set of price and non-price factors. Lastly, while these results
could have arisen as building society depositors may be different from other depositors, it appears that assuming
2
1.z
19
price factors are the only influence on deposit quantities is a strong assumption and that non-price factors such
as choice of deposit accounts are also influential.
Table 5. Model Estimates: Maximum Interest rates
Estimates from Equation 3
Variable GMM System Estimates SUR Estimates
Constant -3.06 (-1.00) -2.92 (-0.99)
Maximum Bank Interest Rate 0.55 (1.22) 0.50 (1.11)
Base Rate 1.22 (4.00)** 1.14 (3.78)**
Mergers 6.87 (2.13)** 6.76 (2.04)**
% Instant Access Deposits 0.077 (1.34) 0.066 (1.25)
Δ deposits*i, t+1 -0.79 (-0.22) -0.66 (-0.30)
ai (0.00) (0.00)
bt (0.00) (0.00)
SE 0.59 0.67
AR(1) (0.48) (0.52)
NORM(2) (0.60) (0.67)
Diff Sargan NA (0.50)
Hausman test NA 99.22 (0.00)
R2
0.36 0.25
Observations 565 565
Estimates from Equation 4
Variable GMM System Estimates SUR Estimates
Constant 7.11 (10.00)** 7.01 (9.80)**
Number of Deposit Accounts 0.15 (2.77)** 0.10 (2.60)**
Number of Branches 0.44 (1.03) 0.38 (0.92)
Δ deposits*i, t+1 0.92 (3.12)** 0.81 (2.69)**
ai (0.00) (0.00)
bt (0.00) (0.00)
SE 0.26 0.32
AR(1) (0.26) (0.22)
NORM(2) (0.29) (0.33)
Diff Sargan NA (0.49)
Hausman test NA 93.45 (0.00)
R2
0.18 0.12
Observations 565 565
Notes: AR(1) is the first order Lagrange Multiplier test for residual serial correlation, undertaken on the
residuals for the SUR estimates and on the first difference of the residuals for the GMM system because of the
first difference transformations involved. SE represents the standard error of the panel estimator. ai and bt are
the fixed and time effects. Sargan tests follow a distribution with r degrees of freedom under the null
hypothesis of valid instruments, for the four instruments in our empirical model. Note: the Difference-Sargan
test is applicable to the GMM system estimator due to the first difference transformations involved. To
establish the validity of the instrument set. NORM(2) is the Jarque-Bera normality test. The endogenous
explanatory variables in the panel are GMM instrumented setting, (.) are p values, (.) are t statistics, *
indicate significant at the 10% level and ** indicates significant at the 5% level
5. Conclusions
This study examines the joint influence of price and non-price factors on changes in retail deposit quantities.
The analysis examines the simultaneous influence of the cost of funds, the average and maximum depository
2
1.z
20
institution specific interest rates, and the number of branches and deposit accounts offered to depositors, on
changes to the quantity of retail deposits held by depository institutions. To achieve this goal, twenty three years
of deposit products, branches, interest rates and deposit quantities for a sample of UK building societies are
examined. The testing procedure was undertaken over two stages in a panel framework to accommodate
concerns with endogeneity of the explanatory variables, contemporaneous correlation of error terms, joint
determination of the variables and potential unknown data dependencies. It is reported both price and non-price
factors are significant influences on deposit quantities, with the former appearing to be marginally more
influential.
It is concluded that when assessing what depository institution actions influence changes in deposit
quantity it is important to persist in using price explanations and acknowledge that non-price factors are also
influential. This finding has importance for both literatures examining risk profiles of depository institution and
the transmission of monetary policy. When considering the risk profile of depository institution, both price and
non-price factors are clearly influential and appear to operate in combination. When considering the
transmission of monetary policy, the interest rate of deposits is not the only factor which affects the quantity of
retail deposits; non-price factors are also influential.
It is recommended these literatures can be profitably advanced in two respects. Further assessment of
bank specific price and non-price factors influencing the eventual deposit quantities needs to be undertaken,
both for different types of bank and internationally. This will enable a clearer picture as to whether these results
are an institutional or national irregularity or are pervasive concerns. In particular examination of other national
markets where varying levels of market concentration may be identified and where data is available and
appropriate, the examination of retail deposits from larger, proprietary banks would be welcome. Second, the
focus of risk assessment of deposit quantity is based on a process which assumes depositors will switch from
deposits to alternative investments. This explanation used widely in prudential regulation would benefit from
further testing and may be a rewarding avenue of future research in light of the historically low levels of interest
observed internationally in recent years.
21
References
Abrantes-Metz, R. M., Kraten, M., Metz, A. D. and Seow, G. S. 2012. Libor manipulation? Journal of Banking
and Finance, 36, no.1: 136-50.
Arellano, M. and Bond, S. R. 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidence and an
Application to Employment Equation, Review of Economic Studies, 58, no.2: 277-97.
Ashton, J. K. 2001. Market Definition in the UK Deposit Savings Account Market. Regional Studies, 35, no.6:
577-90.
Ashton, J. K. 2002. Pricing Behaviour Across Multiple Distribution Channels in the UK Deposit Account
Market: 1997-2001, The International Review of Retail, Distribution and Consumer Research, 12, no.4: 449-
456.
Ashton, J. K. and Hudson, R. S. 2008. Interest rate clustering in UK financial services markets, Journal of
Banking and Finance, 32, no.7: 1393-403.
Ashton, J. K. and Letza, S. 2003. The Differential Returns Offered by Mutually Owned and Proprietary UK
Depository Institutions: 1993-2000. Annals of Public and Cooperative Economics, 74, no.2: 183-204.
Banks, J. and Tanner, S. 1999. Household Saving in the UK, The Institute of Fiscal Studies, London
Basel Committee on Banking Supervision 2010. Basel III: International framework for liquidity risk
measurement, standards and monitoring, Bank for International Settlements, Basel
Biorn, E. and Klette, T. J. 1999. The Labour Input Response to Permanent Changes in Output: An Errors-in-
Variables Analysis Based on Panel Data. Scandinavian Journal of Economics, 101, no.3: 379-404.
Blundell, R. and Bond, S. 1998. Initial Conditions and Moment Restrictions in Dynamic Panel Data Models,
Journal of Econometrics, 87, no.1: 115-143.
Boatwright, P. and Nunes, J. C. 2001. Reducing Assortment: An Attribute-Based Approach, Journal of
Marketing, 65, no.3: 50-65.
Browning, M. and Crossley, T. F. 2001. The Life Cycle Model of Consumption and Saving, Journal of
Economic Perspectives, 15, no.3: 3-22.
Breusch, T. S. and Pagan, A. R. 1980. The Lagrange multiplier test and its applications to model specification in
econometrics, Review of Economic Studies, 47, no.1: 239–253.
Bucks, B. and Pence, K. 2008. Do borrowers know their mortgage terms?, Journal of Urban Economics, 64,
no.2: 218-233.
Building Society Association 2013. BSA Yearbook 2012/13, Publications UK Ltd, Edgeware.
22
Butler, D. and Loomes, G. 1988. Decision Difficulty and Imprecise Preferences, Acta Psychologica, 68, no.1-3:
183-196.
Calem, P. S. and Mester, L. J. 1995. Consumer Behaviour and the Stickiness of Credit-Card Interest Rates, The
American Economic Review, 85, no.5: 1327-1336.
Carbo-Valverde, S., Hannan, T. H. and Rodriguez-Fernandez, F. 2011. Exploiting old customers and attracting
new ones: The case of bank deposit pricing, European Economic Review, 55, no.7: 903-915.
Carlin, B. I and Manso G 2010. Obfuscation, Learning and the Evolution of Investor Sophistication”, Review of
Financial Studies, 24, no.3: 754-85.
Chong, B. C. 2010. Interest rate deregulation: Monetary policy efficacy and rate rigidity, Journal of Banking
and Finance, 34, no.6: 1299-307.
Competition Commission, 2002. The supply of banking services by clearing banks to small and medium-sized
enterprises, 1 Cm 5319, London.
Corvoisier, S. and Gropp, R. 2002. Bank Concentration and Retail Interest Rates. Journal of Banking and
Finance, 26, no.11: 2155-89.
Cornett, M. M., McNutt, J. J., Strahan, P. E. and Tehranian, H. 2011. Liquidity Risk Management and Credit
Supply in the financial crisis, Journal of Financial Economics, 101, no.2: 297-312.
Cremers, K. J. M., Huang, R. and Sautner, Z. 2010. Internal Capital Markets and Corporate Politics in a
Banking Group, Review of Financial Studies. 24, no.2: 358-401.
De Graeve, F., De Jonge, O. and Vander Vennet, R. 2004. The Determinants of Pass-Through of Market
Conditions to Bank Retail Interest Rates in Belgium, National Bank of Belgium Working Paper Series, 47
(Brussels, National Bank of Belgium).
De Haan, L. and Sterken, E. 2005. Asymmetric Price Adjustment in the Dutch Mortgage Market, DNB Working
Paper Series 61 (Amsterdam, De Nederlandsche Bank)
Deaton, A. and Paxson, C. 2000. Growth and Saving among Individuals and Households, Review of Economics
and Statistics, 82, no.2: 212-25.
Entrop, O., Wilkens, M. and Zeisler, A. 2009. Quantifying the Interest Rate Risk of Banks: Assumptions Do
Matter, European Financial Management, 15, no.5: 1001-18.
Financial Services Authority. 2006. Financial Capability in the UK: Establishing a Baseline. The Financial
Services Authority London.
23
Fuertes, A. M. and Heffernan, S. A. 2009. Interest rate transmission in the UK: A comparative analysis across
financial firms and products, International Journal of Finance and Economics, 14, no.1: 45-63.
Fuertes, A. M., Heffernan, S. A. and Kalotychou, E. 2010. How do UK Banks React to Changing Central Bank
Rates? Journal of Financial Services Research, 37, no.2-3: 99-130.
Ghosh, S, and Gregoriou, A. 2008. The Composition of Government Spending and Growth: Is Current or
Capital Spending Better?, Oxford Economic Papers, 60, no.3: 484-516.
Hall, R. E. 1978. Stochastic Implications of the Life Cycle Income Hypothesis, Journal of Political Economy,
86, no.6: 971-87.
Hannan, T. H. and Berger, A. N. 1991. The Rigidity of Prices: Evidence from the Banking Industry, The
American Economic Review, 81, no.4: 938-45.
Heffernan, S. A. 1997. Modelling British Interest Rate Adjustment: An Error Correction Approach, Economica,
64, no.254: 211-31
Heffernan, S. A. 2002. How Do UK Financial Institutions Really Price Their Banking Products?, Journal of
Banking and Finance, 26, no.10: 1997-2016.
Huang, E. and Ratnovski, L. 2011. The Dark Side of Bank Wholesale Funding, Journal of Financial
Intermediation, 20, no.2: 248-63.
Huberman, G. and Jiang, W. 2006. Offering versus Choice in 401(k) Plans: Equity Exposure and Number of
Funds, Journal of Finance, 61, no.2: 763-801.
Hutchison, D. E. and Pennacchi, G. G. 1996. Measuring Rents and Interest Rates Risk in Imperfect Financial
Markets: The Case of Retail Bank Deposits, The Journal of Financial and Quantitative Analysis, 31, no.3:
399-417.
Independent Commission on Banking, 2011. Final Report, London.
Iyerngar, S. S. and Lepper, M. R. 2000. When Choice is Demotivating: Can one Desire Too Much of a Good
Thing? Journal of Personality and Social Psychology, 79, no.6: 995-1006.
Jarrow, R. A. and van Deventer, D. R. 1998. The arbitrage-free valuation and hedging of demand deposits and
credit card loans. Journal of Banking and Finance, 22, no.3: 249-272.
Kirsanova, T. and Sefton, J. 2007. A comparison of national saving rates in the UK, US and Italy”, European
Economic Review, 51, no.8: 1998-2028.
Kalkbrener, M. and Willing, J. 2004. Risk management of non-maturing liabilities. Journal of Banking and
Finance, 28, no.7: 1547-68.
24
Kamenica, E. 2008. Contextual Inference in Markets: On the Informational Content of Product Lines, American
Economic Review, 98, no.5: 2127-49.
Kiser, E. K. 2002. Predicting Household Switching Behavior and Switching Costs and Depository Institutions,
Review of Industrial Organization, 209, no.4: 349-65.
Klemperer, P. 1995. Competition when Consumers have Switching Costs: An Overview with Applications to
Industrial Economics, Macroeconomics and International Trade, Review of Economic Studies, 62, no.4: 515-
39.
Canova, L., Rattazzi, A. M. M. and Webley, P. 2005. The hierarchical structure of saving motives. Journal of
Economic Psychology, 26, no.1: 21–34.
National Audit Office 2009. The Nationalisation of Northern Rock, Report by the Comptroller and Auditor
General, HC 298, The Stationary Office, London
Nyström, K. 2008. On deposit volumes and the valuation of non-maturing liabilities Journal of Economic
Dynamics and Control, 32, no.3: 709-56.
Rosen, R. J. 2002. What Goes UP Must Come Down? Asymmetries and Persistence in Bank Deposit Rates.
Journal of Financial Services Research, 21, no.3: 173-93.
Rosen, R. J. 2007. Banking Market conditions and deposit interest rates, Journal of Banking and Finance, 31,
no.12: 3862-84.
Schmidt, P. 1990. Three-Stage Least Squares with Different Instruments for Different Equations, Journal of
Econometrics, 43, no.3: 389-94.
Shafir, E., Diamond, P. and Tversky, A. 1997. Money Illusion, Quarterly Journal of Economics, 112, no.2: 341-
74.
Sheehan, R. G. 2013. Valuing Core Deposits, Journal of Financial Services Research, 43, no.2: 197-220.
Shin, H. S. 2009. Reflections on Northern Rock: The Bank Run that Heralded the Global Financial Crisis,
Journal of Economic Perspectives, 23, no.1: 101-19.
Staiger, D. and Stock, J. H. 1997. Instrumental Variables Regression with Weak Instruments. Econometrica, 65,
no.3: 557-86.
25
Appendix 1: Sample Institutions
Building Society Start End Building Society Start End
Britannia 1989 2008 Melton Mowbray 1991 2011
Cambridge 1989 2011 National Counties** 1990 2011
Chelsea 1990 2008 Nationwide 1989 2011
Coventry 1989 2011 Newbury 1990 2011
Darlington 1990 2011 Newcastle 1989 2011
Furness 1991 2011 Norwich and Peterborough 1989 2008
Hanley Economic* 1991 2011 Principality 1989 2011
Hinckley and Rugby 1990 2011 Progressive 1989 2011
Ipswich 1991 2011 Saffron 1990 2011
Leeds and Holbeck 1989 2011 Skipton 1989 2011
Leek United 1990 2011 Stroud and Swindon 1989 2008
Market Harborough 1991 2011 Teachers' 1990 2011
Marsden 1990 2011 West Bromwich 1989 2011
* One year where no instant access accounts offered (2002)
** Two years where no instant access deposits were offered (1999-2000)